Iterative process for optimizing optical character recognition

• US 9,870,520 B1
• Filed: 08/02/2013
• Issued: 01/16/2018
• Est. Priority Date: 08/02/2013
• Status: Active Grant

• Current Assignee: Intuit, Inc.
• Original Assignee: Intuit, Inc.
• Inventors: Becker, Richard J., Chan, Horace, Sikri, Neha, Moise, Daniel L.
• Primary Examiner(s): Rush, Eric
• Application Number:
  US13/957,709
• Time in Patent Office: 1,628 Days
• Field of Search: 382309, 382181, 382310, 382321, 382100, 382101, 382105, 382112, 382135-140, 382182, 382311
• CPC Class Codes: G06F 18/21 Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
• CPC Class Codes: G06F 18/217 Validation; Performance evaluation; Active pattern learning techniques
• CPC Class Codes: G06V 10/98 Detection or correction of errors, e.g. by rescanning the pattern or by human intervention; Evaluation of the quality of the acquired patterns
• CPC Class Codes: G06V 30/226 of cursive writing
• CPC Class Codes: G06V 30/2528 Combination of methods, e.g. classifiers, working on the same input data
• CPC Class Codes: G07B 2017/00725 Reading symbols, e.g. OCR

Abstract

The disclosed embodiments relate to a system and method for calibrating optical character recognition (OCR) processes for an image captured through a mobile computing device. During operation, the system adjusts the OCR process through pre-recognition functions, OCR functions and/or post-recognition functions with multiple sets of parameter settings. With each of these sets, the system scores the OCR process output against an image with known text. Once the sets are scored, the system sorts the sets of parameters, removes some sets, then mixes and mutates the remaining sets in a process akin to evolutionary biology. By repeating this procedure, the system produces a set of parameter settings that can be used to calibrate OCR processing.

Claims

1. A computer-implemented method for optimizing parameter settings for an optical character recognition process, comprising:

receiving a population comprising at least two sets of parameter settings;

receiving an image comprising known text;

for each set of parameter settings in the population, determining an accuracy score by;

executing a set of pre-recognition functions using the set of parameter settings to modify the image according to the parameter settings to produce a modified image;

executing an optical character recognition function on the known text in the modified image to produce an output; and

testing the output to determine the accuracy score for the set of parameter settings by comparing the known text to the output;

selecting a first set of parameter settings from the population based on the determined accuracy scores;

receiving a second image separate from the image;

executing the set of pre-recognition functions using the first set of parameter settings to modify the second image according to the first set of parameter settings to produce a modified second image; and

executing the optical character recognition function on the modified second image to produce a second output.

2. The method of claim 1, wherein after determining an accuracy score for each set of parameter settings in the population, the method further comprises:

iteratively performing the following operations;

removing at least one set of parameter settings from the population based on the accuracy scores; and

mixing the parameter settings from the remaining sets of parameter settings to form new sets of parameter settings in the population.

3. The method of claim 2, wherein after mixing the parameter settings to form new sets of parameter settings, the method further comprises randomly changing at least one parameter setting in at least one set of parameter settings in the population.

4. The method of claim 2, wherein the operations are iteratively performed until at least one of the accuracy scores exceeds a predetermined threshold.

5. The method of claim 1, wherein executing the optical character recognition function involves using the set of parameter settings.

6. The method of claim 1, wherein after executing the optical character recognition function, the method further comprises executing a set of post-recognition functions using the set of parameter settings to modify the output.

7. The method of claim 1, wherein the method further comprises adjusting at least one of the sets of parameter settings based on environmental factors associated with acquisition of the image.

8. The method of claim 1, wherein receiving the image comprising known text comprises receiving the image comprising known text in a first environment, wherein receiving the second image separate from the image comprises receiving the second image in the first environment, and further comprising:

receiving a third image comprising known text in a second environment;

for each set of parameter settings in the population, determining a second accuracy score by;

executing the set of pre-recognition functions using the set of parameter settings to modify the third image according to the parameter settings to produce a modified third image;

executing the optical character recognition function on the known text in the modified third image to produce a third output; and

testing the third output to determine the second accuracy score for the set of parameter settings by comparing the known text to the third output;

selecting a second set of parameter settings from the population based on the determined second accuracy scores;

receiving a fourth image separate from the third image in the second environment;

executing the set of pre-recognition functions using the second set of parameter settings to modify the fourth image according to the second set of parameter settings to produce a modified fourth image; and

executing the optical character recognition function on the modified fourth image to produce a fourth output.

9. A non-transitory computer-program product for use in conjunction with a computer system, the computer-program product comprising a computer-readable storage medium and a computer-program mechanism embedded therein, to optimize parameter settings for optical character recognition, the computer-program mechanism including:

instructions for receiving a population comprising at least two sets of parameter settings;

instructions for receiving an image comprising known text;

for each set of parameter settings in the population, instructions for determining an accuracy score by;

executing a set of pre-recognition functions using the set of parameter settings to modify the image according to the parameter settings to produce a modified image;

executing an optical character recognition function on the known text in the modified image to produce an output; and

testing the output to determine the accuracy score for the set of parameter settings by comparing the known text to the output;

instructions for selecting a first set of parameter settings from the population based on the determined accuracy scores;

instructions for receiving a second image separate from the image;

instructions for executing the set of pre-recognition functions using the first set of parameter settings to modify the second image according to the first set of parameter settings to produce a modified second image; and

instructions for executing the optical character recognition function on the modified second image to produce a second output.

10. The computer-program product of claim 9, further comprising instructions for iteratively performing the following operations after determining an accuracy score for each set of parameter settings in the population:

removing at least one set of parameter settings from the population based on the accuracy scores; and

mixing the parameter settings of the remaining sets of parameter settings to form new sets of parameter settings in the population.

11. The computer-program product of claim 10, further comprising instructions for randomly changing at least one parameter setting in at least one set of parameter settings in the population after mixing the parameter settings to form new sets of parameter settings.

12. The computer-program product of claim 10, wherein the instructions are configured to be iteratively performed until at least one of the accuracy scores exceeds a predetermined threshold.

13. The computer-program product of claim 9, wherein the instructions for executing the optical character recognition function are configured to use the set of parameter settings.

14. The computer-program product of claim 9, further comprising instructions for executing a set of post-recognition functions using the set of parameter settings to modify the output after executing the optical character recognition function.

15. The computer-program product of claim 9, wherein the computer-program product further includes instructions for adjusting at least one of the sets of parameter settings based on environmental factors associated with acquisition of the image.

16. A computer system, comprising:

a processor;

a memory; and

a program module, wherein the program module is stored in the memory and configurable to be executed by the processor to optimize parameter settings for optical character recognition, the program module including;

instructions for receiving a population comprising at least two sets of parameter settings;

instructions for receiving an image comprising known text;

for each set of parameter settings in the population, instructions for determining an accuracy score by;

executing a set of pre-recognition functions using the set of parameter settings to modify the image according to the parameter settings to produce a modified image;

executing an optical character recognition function on the known text in the modified image to produce an output; and

testing the output to determine the accuracy score for the set of parameter settings by comparing the known text to the output;

instructions for selecting a first set of parameter settings from the population based on the determined accuracy scores;

instructions for receiving a second image separate from the image;

instructions for executing the set of pre-recognition functions using the first set of parameter settings to modify the second image according to the first set of parameter settings to produce a modified second image; and

instructions for executing the optical character recognition function on the modified second image to produce a second output.

17. The computer system of claim 16, further comprising instructions for iteratively performing the following operations after determining an accuracy score for each set of parameter settings in the population:

removing at least one set of parameter settings from the population based on the accuracy scores; and

mixing the parameter settings of the remaining sets of parameter settings to form new sets of parameter settings in the population.

18. The computer system of claim 17, further comprising instructions for randomly changing at least one parameter setting in at least one set of parameter settings in the population after mixing the parameter settings to form new sets of parameter settings.

19. The computer system of claim 17, wherein the instructions are configured to be iteratively performed until at least one of the accuracy scores exceeds a pre-determined threshold.

20. The computer system of claim 16, wherein the instructions for executing the optical character recognition function are configured to use the set of parameter settings.

21. The computer system of claim 16, further comprising instructions for executing a set of post-recognition functions using the set of parameter settings to modify the output after executing the optical character recognition function.

22. The computer system of claim 16, wherein the program module further includes instructions for adjusting at least one of the sets of parameter settings based on environmental factors associated with acquisition of the image.